[bitbake-devel] [PATCHv2] fetch2/gitsm: copy submodules from DL_DIR in unpack

[Christopher Larson]

On Wed, Aug 23, 2017 at 8:01 AM, Christopher Larson <kergoth at gmail.com>
wrote:

> On Wed, Aug 23, 2017 at 3:22 AM, Richard Purdie <richard.purdie@
> linuxfoundation.org> wrote:
>
>> On Mon, 2017-08-21 at 21:57 +0500, Christopher Larson wrote:
>> > From: Christopher Larson <chris_larson at mentor.com>
>> >
>> > Revert "fetch2/gitsm: Fix when repository change submodules"
>> >
>> > This reverts commit 12f6c0651af8bd5d6efb751690571cf2fcd3eeb0.
>>
> […]
>
>> 2017-08-23 11:20:14,819 - oe-selftest - INFO - Ran 1 test in 33.385s
>> 2017-08-23 11:20:14,819 - oe-selftest - INFO - FAILED
>> 2017-08-23 11:20:14,819 - oe-selftest - INFO -  (failures=1)
>> 2017-08-23 11:20:14,824 - oe-selftest - INFO - RESULTS:
>> 2017-08-23 11:20:14,824 - oe-selftest - INFO - RESULTS -
>> devtool.DevtoolTests.test_devtool_add_fetch_git - Testcase 1161: FAILED
>> 2017-08-23 11:20:14,824 - oe-selftest - INFO - SUMMARY:
>> 2017-08-23 11:20:14,824 - oe-selftest - INFO - oe-selftest () - Ran 1
>> test in 33.391s
>> 2017-08-23 11:20:14,824 - oe-selftest - INFO - oe-selftest - FAIL -
>> Required tests failed
>>
>> Could you take a look at that test case and see what the issue is please?
>>
>
> Will do, thanks, sorry about that. I ran the bitbake selftests, but didn’t
> think to re-run the oe selftests. *eyeroll*
>


This is uglier than I had been thinking. If the upstream url for an
existing module changes between srcrevs or branches, we’re in trouble,
since the modules/ dir in clonedir won’t line up with .gitmodules. I was
thinking about adding a needs_update() to GitSM to check for the case where
it doesn’t match up, but then if multiple recipes need different sets of
modules for the same clonedir at the same time, they’d both run download()
and first one would switch to one configuration, then the other would
switch back..

The perils of using a single directory in a single clone for multiple
possible sets of modules and revs.. I don’t think it’s worth trying to
handle this with a stopgap measure, best to just punt on this for this
release and pursue the rewrite of gitsm after the release. Feel free to
drop these patches. Thanks.
-- 
Christopher Larson
kergoth at gmail dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Senior Software Engineer, Mentor Graphics
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/bitbake-devel/attachments/20170823/6be6759e/attachment-0002.html>